The pixel-level adaptive gain imaging system can achieve a large dynamic range of remote sensing imaging requirements while ensuring high signal-to-noise ratio by integrating four different volume integrating capacitors in the electronic link of each pixel. During prelaunch testing of this imaging system, due to the large dynamic range of Ultra Low Gain (ULG) and the limited energy of the laboratory integrating sphere, the output characteristics of the second half range of ULG can only be indirectly calibrated by recurrence of the proportional coefficient with Low Gain (LG). Excessive reflected energy from the onboard solar scaler can lead to saturation of High Gain (HG) and Medium Gain (MG) outputs, which can only be inferred through proportional coefficients and cannot be directly calibrated. A scheme for measuring the onboard proportion coefficient between adjacent gains is proposed. The test images are classified using outputs of different gains as features to obtain the output of different imaging targets, and the outputs of different targets are used as multiple calibration energy levels. The proportion coefficient of adjacent gains is obtained by linear fitting of adjacent gain outputs using the least squares method. This scheme verifies the laboratory proportionality coefficient and solves the problem of onboard radiance calibration of HG and MG.The onboard gain ratio measurement scheme of the pixel-level adaptive gain imaging system proposed in this article mainly utilizes the incoming pupil radiance received by the detector to provide appropriate energy levels for adjacent two gain levels, in order to calculate the conversion ratio coefficient between them. During the calibration process, the type of imaging target is not taken into account, that is, the ground, clouds, etc. which are considered as energy levels. Therefore, atmospheric correction and cloud removal are not required, and it is not affected by weather. This scheme can be used to measure adjacent gain ratio coefficients on both sunny and cloudy days. The specific solution is to determine the two level gain ratio coefficients that are currently suitable for measurement through the adaptive gain imaging results of different channels; Then, the fourth gear gain output after removing stripe noise is used as the imaging feature of a single pixel to cluster the imaging images; By using different categories of image regions with adjacent gains after clustering as energy points to fit the linear relationship between adjacent gains, the ratio coefficient of adjacent gains for the current channel can be obtained. And further use the gain ratio coefficient to calculate the onboard radiation calibration coefficients for the four gains.The removal effect of stripe noise in the imaging process of outfield experiments and the complexity of ground targets during outfield imaging may have a certain impact on the measurement accuracy of the gain to gain ratio coefficient. The proportion coefficient calculated using the method proposed in this article is used to invert the higher gain image from the lower gain image in the adjacent two levels. Compared with the actual higher gain image, the normalized mean square error is mostly less than 0.01, the structural correlation coefficient of the two images is basically around 90%, and the data correlation coefficient reaches 90%. Prove that this method has high accuracy in determining the correlation coefficient between adjacent two gain ratios, and it is highly feasible to use it for recursive calculation of high gain radiometric calibration coefficients in satellite radiometric calibration.Through field imaging experiments, it has been verified that the ratio coefficient of adjacent gains measured by this method can effectively complete the task of inverting high gain images from low gain images. Therefore, it is highly feasible to use this method to obtain the other three gain radiometric calibration coefficients from the radiometric calibration coefficients of ULG during onboard radiometric calibration. This provides a certain idea for the onboard radiation calibration of large dynamic range imaging systems with multiple gains.